RM Flashcards
(149 cards)
1
Q
general
Differences IC and ordinary crimes
A
(1) Criminalized by treaty, not by national government
(2) crimes of large scale, committed under particular conditions (during conflict).
(3) evidence: OC more focus on forensic evidence by contrast with IC trials
2
Q
The tesseras criminologica of IC
A
- Prevalence: Measurements and occurrence of crime. Predominantly quantitative.
- Etiology: Causes of crime. Explains crime with bio-/psychological theories, the situational characteristics of crime (socioeconomic, cultural or geophysical make-up), integrated/macro theories.
- (Non)-judicial responses: Legal or extralegal responses (of sentencing). Differences in investigation and trial.
- Victims: Victim studies which focus on its characteristics, well-being, aftermath. In IC focus on societal impact, such as socio-economic impact, health consequences (famine), or ecological impact.
3
Q
Quantitative –> analytical empirical
A
- Inductive phase: Start with research idea into RQ
- Deductive phase: how we are going to measure the construct to answer the RQ (conceptualization, operalization)
- Data collection
- Analysis
- Evaluation
4
Q
Qualitative –> interpretative empirical: features
A
Grounded theory = The methodology involves the construction of hypotheses and theories through the collecting and analysis of data. Grounded theory involves the application of inductive reasoning.
Conceptualization is more inductively, generally working from empirical data as they emerge.
5
Q
Saturation
A
By the time the explanation or interpretation of the phenomenon under study converges in the sense that the explanation does not change anymore upon collection of newdata, the research terminates.
6
Q
Narrative view
A
= Give an overview and summary of a number of relevant studies that have been published.
7
Q
Systematic review
A
= Aim to identify all relevant studies through a systematic search with keywords across multiple databases.
8
Q
Meta-analysis
A
= A systematic review where not only a summary is given, but the data from all previous studies are combined into a new aggregated dataset and re-analyzed.
9
Q
Construct validity
A
= To what extent does an empirical measure reflect the real meaning of the concept?
For an instrument to have construct validity it first has to meet all the other validities
10
Q
Content validity
A
= Degree to which a measure covers the range of characteristics included within a concept.
11
Q
Criterion validity
A
= The scores obtained with the instrument should correlate with an external criterion that you would expect it to correlate with.
12
Q
Construct validity-in-the-narrow-sense
A
= Whether a measure is built in such a way that does not discriminate against certain research subjects → understanding of subjects.
13
Q
Other validities
A
Face validity, statistical conclusion validity, external validity
14
Q
Internal validity
A
= The degree of confidence that the causal relationship you are testing is not influenced by other factors or variables. Confounders or third variables can generate an association between two properties that are not causally related.
15
Q
Reliability
A
= Do we have precise measurements? Do identical measurements end up in identical results? Refers to the precision with which a construct is measured.
16
Q
Ethics: 6 features
A
- Decent and respectful treatment of research subjects or respondents.
- Informed consent for the respondents.
- Ensure the safety of respondents.
- Do no harm principle = Entails that as a researcher it is one’s duty to minimize the risk that research participants suffer adverse consequences from participating in the research
- Confidentiality, anonimousity for respondents and their data.
- Data should be transported safely and stored securely.
17
Q
Populations and samples connotation:
A
We write sample characteristics with Latin literals (M, s, rXY ); whenever we refer to properties of the population we use Greek literals (such as µ for the mean.
18
Q
Aim sampling
A
- Quantitative studies: Representativeness & generalizability.
- Qualitative studies: Maximizing information & saturation.
19
Q
Litmus test
A
= When every population member has an equal chance to end up in the sample.
20
Q
Random sample: def, terms (2), pros, cons
A
Drawn from a complete pre-existing sampling frame.
Sampling frame = All persons who have a chance to be included into the sample (e.g., a list, area sampling).
Sampling error = Deviations that occur by chance.
Pros: ensures external validity
Limits: list of people not always available
21
Q
Systematic sample
A
= Type of probability sampling method in which sample members from a larger population are selected according to a random starting point but with a fixed, periodic interval.
Ensures spread (not distribution). Is efficiënt when you e.g. don’t have a framework to work with.
22
Q
Stratified sample
A
= Researchers divide subjects into subgroups called strata based on characteristics that they share (e.g., race, gender, educational attainment). Once divided, each subgroup is randomly sampled using another probability sampling method.
Advantage: Ensures representation of relevant strata/equal precision over strata (distribution).
Disproportionate stratified sampling = Allows the researcher to give a larger representation to one or more subgroups to avoid underrepresentation of the said strata. This applies to populations with a very high strata population ratio.
23
Q
Cluster sample
A
= Divide population into smaller random groups in clusters. Then randomly select among clusters to form a sample. Very efficient for large populations. The more clusters with small groups → higher validity.
Limitations:
- Sampling error –> more noisy clusters
- DEFF
Multistage cluster sampling = Rather than collect data from every single unit in the selected clusters, randomly select individual units from within the clusters.
24
Q
DEFF
A
Design effect. Estimates based on a cluster sample tend to be more volatile than estimates based on a flat random sample. This is due to the fact that cluster members resemble each other. The more respondents within a cluster, the larger the cluster, the larger DEFF.
Rule of thumb: 1-3 → above 3 is problematic. and results could be too noisy.
If DEFF is 3 🡪 the variability is thrice that of a random sample of the same size. If DEFF is 1.40 🡪 this means the variance is 40% increased. As DEFF is ratio, a DEFF of 1 would indicate no difference.
25
Convenience sample
One simply selects for the sample whomever is available. Useful for exploratory studies → no representativeness.
26
Quota sample
Like a stratified sample in the sense that per stratum a desired number of sample members is drawn: the researcher attempts to guarantee that the relative representation of certain properties of sample members is as it is in the population. However, relies on the non-random selection of a predetermined number or proportion of units.
Difference with strata: Quota way less rigorous → no random sampling in quota.
27
Purposive sample
= Sample members are selected due to their characteristics and targeted because the researcher expects them to have relevant information (key informants) → not generalizable.
Often used in ethnographic research.
28
Snowball sample: def, issues (4)
One starts with a certain number of accessible respondents. Favorable for hidden populations.
Disadvantage:
a. Lengthy sampling,
b. No reliable respondents
c. Quality dependent on the zero-stage respondent
d. Not generalizable.
29
Respondent-driven sampling
= In which a respondent receives coupons that s/he can distribute among prospective new respondents → bonus system.
30
Indirect sampling
= One select respondents via key informants (= the informant method). Indirect sample as part of snowball and/or combination with purposive sampling.
31
Expanded Programme on Immunization (EPI): def + issue
= Selecting clusters proportional to size + spinning a pen. Commonly used when the sampling frame is unknown. For very remote locations.
Main-street biased = EPI is not random if it starts at the center (center-biased). Researchers start with main street, so not using periphery. Main street households are more likely to be affected (e.g. shootings, police stations located etc.).
32
Noise and bias
Noise = The sample members’ averages **vary** a lot around the true population score that we are interested in. On average combined over all samples, they may provide a good estimate of the population mean. 🡪 can still be representative
Bias = The sample averages then **differ systematically** from the population mean 🡪 not representative.
33
(Non) response terms
Sample nonresponse = Not all those we establish contact with, will consent to participate.
Retention rate = Completed respondents.
Attrition rate = Not-completed respondents.
34
Ethnographic interview
= Spontaneous and informative conversation.
35
Respondent overview
= Formal interview with an official role for the researcher.
36
Informant interview
= Focus on a group of people and not the respondent.
37
Interview options
1. Open-ended interview
2. Topical interview
3. Questionnaire interview
4. self-administered questionnaire
38
Informed consent
= Entails an agreement (‘consent’) to partake in the study after full information (‘informed’) has been received of the research.
39
Open-ended interviews: def and types
= Unstructured. In-depth interview. Aim is to understand.
Types:
a. ***Narrative interviews*** = Unstructured interviews where respondents are encouraged to relate their experiences or tell stories rather than respond to questions.
b. ***Oral history*** = To reconstruct events of the past (for eyewitnesses etc.)
c. ***Biographical interview/life story interview*** = Focus on recounting events of the past of the interviewee’s own life.
40
Open-ended interviews/ethnology: analysing
Field notes = Extensive summaries away from respondent or observation.
Analytic memo’s = Analysis generally starts immediately after interviews or observation has taken place; researchers write analytic memos, in which they describe everything that occurred, what patterns emerge in the data, what hypotheses are formed and rejected, what concepts best synthesize the findings.
41
Topical interview: def and terms (2)
Semi-structured with a topic list. Could also be pre-set list with open-ended questions.
Intimacy curve = Build up and off the sensitivity.
Focus group = Good fit for focus group.
Diverging views or opinions of the group members → efficient way to get an overview of viewpoints.
42
Questionnaire interviews: def and advantages (2)
= Quantitative, fixed and structured. Multiple-choice.
Advantage:
(1) Standardization serves in fact to minimize distortion through interviewer behavior or wording of the specific questions,
(2) minimizes the influence of the researcher on the interviewee.
43
Self-administered questionnaire: pros (1), cons (3)
Advantage: Financial advantage.
Disadvantage: (1) no room for explaining, (2) respondents more likely to give up, (3) minimum literacy required.
44
Self-administered questionnaire: use of computers (4 advantages)
Advantages:
a. Automatization of things like errors, chronological order etc. → **less missing values**.
b. **Better validity:** Respondents on sensitive topics.
c. Data entry is immediate, and thus prevents mistakes with data entry by hand.
d. Promotes confidentiality and privacy → **prevents social desirability bias.**
45
LAAF procedure
= Life as a Film. In the LAAF elicitation procedure, respondents are asked to describe their life as a movie.
46
Life History Calendar
= Respondents are asked to fill out their lives along a timeline, with important events pointed out.
47
Types of questions (3)
Generative questions = Get the respondent to talk.
Directive questions = Closed-end questions. This is to elicit reactions.
Screener questions = Whether the respondent was ever the victim of a burglar
48
agreement response bias
(= tendency to agree with every statement)
49
2 classes of measurements:
1. Measures of central tendency: Descriptive: mean (only for interval level up), median (only for ordinal level up), mode.
2. Measures of variability: Variance and SD (only for interval level up).
50
Describing the distribution of scores for variables lower than interval level:
(1) Frequency tabulation
(2) pie/bar chart
51
Bivariate analysis
= The interrelation of two variables.
52
Odds ratio
Tells us something about the strength of the **association** between two dichotomous variables. Expresses this association in terms of risk.
53
Difference regression model and ANOVA:
Regression and the analysis of variance model is that in the regression case the independent variables are of interval measurement level or higher, while in the analysis of variance model they are nominal.
54
Synonyms independent variables
exogenous variables, predictors, input variables, explanatory variables.
55
Synonyms dependent variables
endogenous variable, output variable, outcome.
56
Mediating variables
= The independent variables are themselves **also predicted by other independent variables**; they may be called mediating variables.
57
Parsimony
= We prefer our models to describe data as well as possible. At the same time, we want them to be as simple as possible → principle of parsimony.
58
ANOVA: def, testing, H0, w2
= Analysis of variance. Shows whether categories of independent variable(s) help predict Y, based on analysis of means. Kijktof de populatiegemiddelden van meer dan 2 groepen van elkaar verschillen.
Level of measurement: Categories of independent variables are nominal variables. Y is continuous.
Testing: By using F-distribution --> indicated by p value
H0: No differences between the levels of the independent variable.
ω2 tells you how much of the variance of the dependent variable is explained by the entire ANOVA model → same as R2
59
Similarities between regression analysis and ANOVA:
In each case we are trying to predict a dependent variable from one or more independent variables. In each case the model is linear.
60
Theoretical sampling
= Process, where one samples respondents up to the point where adding new sample members would not change the conclusions.
61
Theoretical validity
= Saturation has been reached, and theoretical validity achieved.
62
First level coding
= Here one essentially tries to give labels to the content of the data, tries to group sentences or pieces of text under meaningful headers.
63
Second level coding
= After the raw material has been coded, at the stage of second level coding one starts analyzing the data more thoroughly.
64
Content analysis
= Aims to assess the meanings or symbolic or covert intent of communications. Wide class of analytic techniques that attempt to uncover the hidden or implicit meanings of communications. Both quantitative and qualitative approaches!
65
Verbatim
= Respondents formulate precisely, or a life-story of a respondent in a sense illustrates the themes so well, that these are verbatim reproduced.
66
Setting of criminological research: general issues (10)
a. Offenders: Offenders will likely not report certain offending behavior. They have no common interest with the researchers. They have shame, mental illness, drug abuse, impaired memory.
b. Interval validity issue: There is no possibility of chance administering intervention X or control intervention → prosecution and judge are responsible for that.
c. Sample issues: Offenders are often poor, unregistered and difficult to reach.
d. Victims: Afraid of being re-traumatized.
e. Serious crimes are rare events.
67
Setting of criminological research: statistical issues (10)
1. Samples are not representative as criminologists lack sampling frames.
2. Small number of offenders in population, small number of offenders responsible for large proportion of all offending.
68
Setting of criminological research: methodological issues (10)
1. Dark number → police data is biased.
2. Non-experimental designs → researcher as bystander. Judge is the one who intervenes.
3. Policy and political relevance makes it highly politicized.
69
Double dark number (12)
= When authorities are involved in conflict they do not attempt to uncover crimes or maybe actively hide crimes.
70
Excess mortality
= Term frequently encountered in prevalence studies. Excess mortality is defined as mortality exceeding ordinary mortality levels. Excess mortality of civilians directly attributed to violence → might indicate criminal attacks.
71
Baseline mortality
= Normal mortality expressed using two epidemiological measures: in the crude mortality rate (CMR) and in the under 5 mortality rate (U5MR)
72
CMR
= Defined as the total number of deaths that occurred in a population of a certain size during a certain period.
Direct mortality to estimate overall mortality in a population and calcullate excess deaths (if baseline is known).
died/10,000/day OR #died/1,000/year
73
U5MR
= Defined as the total number of deaths of children under 5 years of age that occurred in a population of children under 5 years during a certain period.
Indirect mortality to calculate excess deaths (if baseline is known).
children<5 died/x <5 population /times unit
# died/10,000/day #died/1,000/year
74
Methods for assessing mortality:
1. Counting
2. Estimating
75
Methods for assessing mortality: Counting: preconditions
Preconditions:
1. availability of good population administration systems before conflict
2. good administration systems during and after conflict.
76
Methods for assessing mortality: Estimating:(4)
Methods:
1. **Retrospective surveys** = Interviewing family members of the deceased.
- Sampling: Cluster sampling with EPI
- Survivor/mortality bias
2. **Combining** data from existing mortality surveys with other data
3. **Counterfactual** methods that use **extrapolations**
4. **Capture-recapture** methods
77
Aim etiology
Aimed at understanding the individual experiences, motivations, and societal, familial, ideological influences on the behaviour.
78
Etiology: micro-level (+term)
How can we design studies that can help identify those properties and situational characteristics that distinguish the perpetrators from non-perpetrators and that may validly serve as explanation for participation in such crimes?
Symbolic realism = The need for social researchers to treat the beliefs they study as worthy of respect rather than as objects of ridicule.
79
Etiology: meso-level
Looking for patterns, coming together at the group or organization level. Structural level (group psychology etc.).
Atrocity producing situations: institutionalized dehumanization of the victimized group + extreme power differences, open opportunity structures, lack of social control
Ex: prisons, concentration camps, death camps, but also armed conflicts in general.
80
Etiology: macro-level
Investigate global or country-level causes of conflict. Supranational level, country-level strife going back to colonial times or economic downturns.
Prone factors for conflict: (a) difficult life conditions, (b) exclusionary ideology, © ethnic cleavages, (d) prior conflict, (e) authoritarian regime, (f) isolated international politics.
81
Proxy variable
= Variable representing but not measuring a relevant variable (e.g. GDP for standard of living).
82
Etiology: methodological issues (3)
1. In searching for explanations for IC → mechanisms (manner of conduct etc.) may **evolve** during the conflict.
2. **Concurrent explanation**: Not all conflicts are equal, different factors may explain different conflicts → no comprehensive all-in-all explanation. The etiology theories co-exist.
3. **Retrospectiveness issue**: Often no resource to ask perpetrators why, in hindsight, they committed their crimes.
83
Interviewing detained persons: benefits (4)
1. Prisoners are easy to find. Efficient search algorithm.
2. One may assume that once people have been convicted they will have little reason to lie about their acts.
3. Prisoners have so much free time and are generally bored, and will therefore likely agree to participate.
4. ‘Cheap’: as they are not employed and most prisons forbid handing out gifts, thus for free.
84
Interviewing detained persons: issues (3)
1. Not every perpetrator will be apprehended → select group.
2. Unclear whether prisoners can really speak freely: prisons are not always extremely private places.
3. Control prisoners: If prison wardens or directors decide who may be interviewed and who not, there is little control over sample selection.
85
# general?
Interviewing perpetrators: tips (8)
1. Building up **good rapport**, often over a lengthy period
2. Minimizing the investigation of more **sensitive topics** the informant might be involved in, to reduce incentives to fabrication
3. Using **internal triangulation**, by interviewing different kinds of perpetrators
4. Judging the frankness of an informant by his/her willingness to **accept objective facts** about the conflict.
5. Not interviewing informants with realistic reasons to **fear prosecution**.
6. Offering the assurance of complete **anonymity**.
7. **Covering given topics in multiple ways**, including repeated interviewing, and visiting locations to verify or revisit accounts.
8. **Cross-referencing** of findings to other qualitative or quantitative data, collected by different researchers.
86
Special responses
= Responses to IC that do not emanate from strictly judicial institutions, but that nevertheless may give rights and have legal implications (Ex: amnesties).
87
Methodological issues when studying reactions to IC: (4)
1. **Small samples**: When studying legal responses to IC by tribunals and the ICC, often small samples.
2. **Assessing impact post-conflict interventions**: Difficult to assess the effects of amnesties and truth commissions for lack of adequate ‘controls’. No control cases are available.
3. **No empirical experiments**:
4. **Confounding variables**: Lack of adequate controls of these variables → need for qualitative methods that are able to investigate the workings of a multitude of interrelated factors in specific historical and geopolitical contexts.
88
Methods for assessing societal consequences IC: (5)
Intergenerational effects
Economic damage
Effects on schooling
Demographic consequences
Health consequences
89
Scientific inquiry
= **Organized** and **systematic** as opposed to theoretical or philosophical reasoning. **Empirical** method for acquiring knowledge.
90
Characteristics empirical research: (5)
(1) systematic observation, (2) measurement, (3) critical thinking/skepticism, (4) hypothesis testing, (5) deriving findings from data/observations.
91
Research aims: general (4)
exploration, description, explanation, evaluation
92
Indicators and dimensions in IC
(1) Indicators: An observation that we choose to consider as a **reflection** of a variable we wish to study (e.g. cessation of conflict, increased court case load etc., )
(2) Specifiable **aspect** of a concept (e.g. peace dimension, rule of law dimension).
93
Progression of measurement: (4)
(1) conceptualization, (2) nominal definition, (3) operational definition, (4) measurements in the real world.
94
Obtrusive research methods: Primary data
2 types:
1. **Interviews, survey research, focus groups**: Deciding on (a) structure, (b) conducting interview, (3) recording, (4) analyzing.
2**. Observation, field research, experiments:** Observer has various *degrees of participation* (full participant - complete observer) and from *active inquiry* (interviewing) to passive observation.
95
Issue of reactivity
= Subjects of research may react to the fact of being studied and alter their behavior.
--> in observations/field research
96
Unobtrusive research: Secondary research sources (4)
Sources: (1) Content analysis, (2) analysis of secondary data, (3) literature review, (4) meta-analyses.
97
Mixed methods designs: (3)
1. Sequential Quan → Qual design = Quantitative and qualitative strands in sequence with the purpose of using follow-up qualitative data to **elaborate, explain,** or confirm initial quantitative results.
2. Sequential Qual → Quan design = Qualitative and quantitative strands in sequence with the purpose of using follow-up quantitative data to **generalize, test,** or confirm initial qualitative results.
3. Parallel / convergent / concurrent design = Qualitative and quantitative simultaneous → integrated findings for **complementary and more complete and validated** conclusions.
98
Qualitative assesment goals:
1. Reliability
2. Validty
3. Precision and accuracy
99
Qualitative trustworthiness goals (5)
1. Validity
2. Reliability:
3. Reflexivity: Role of the researcher is considered.
4. Confirmability: The extent to which findings are qualitatively confirmable through the analysis being grounded in the data. Neutrality, objectivity. Clear link between data and findings
5. Transferability
100
Qualitative assesment goals: reliability
Accuracy of the measurement procedures to consistently produce the same scores → replicability.
Techniques to obtain reliability: (1) test-retest method (over time), (2) using established measures, (3) inter-rater reliability.
101
Qualitative assesment goals: validity
Types:
1. Internal validity: Degree of confidence in the findings of a study, whether alternative explanations are possible.
2. External validity
3. Construct / measure validity = To what extent does an empirical measure reflect the real meaning of the concept? measure with down below
Face validity
Content validity
Criterion validity
Construct in the narrow sense (discriminant) validity
102
Qualitative assesment goals: precision and accuracy
+- Reliability.
Improving precision improves reliability of the methods.
Accuracy is concerned with whether the response is a correct reflection of the real world.
103
Differences probability vs non-probability sample:
1. Budget: Large budget → random. Smaller budget → cluster or systematic sampling.
2. Population characteristics: Homogeneous → cluster sampling. Heterogeneous → stratified sampling.
3. Prior knowledge: Available → stratified sample. Unavailable → cluster or systematic sampling.
104
Inferential approaches
= Objective is to infer from a sample to a population (quant) or from your findings to a phenomenon/theory (qual). Focus on explanation, not just description.
105
Descriptive statistics: (3)
1. Mean, median (middle value), mode (common value).
2. Variance s2: Spread between numbers in a dataset. How far each number is from the mean.
- De variantie is de gemiddelde kwadratische afwijking ten opzichte van het gemiddelde.
3. Standard deviation (SD): square root of variance: dispersion of a dataset relative to its mean.
- SD is de gemiddelde afwijking hoever een score zit van het gemiddelde.
106
Bivariate analysis: types
1. Correlation (pearson, spearman, kendall)
2. Chi square x2
3. Odds Ratio
4. Simple linear regression
5. ANOVA
107
Correlation: def, types, rule of thumb
**Indicates direction and strength** between two numerical variables.
1. **Pearson r**: Continuous variables (interval, ratio), normal distribution.
2. **Spearman’s r:** For ordinal variables or not-normally distributed continuous variables.
Correlation R: Rule of thumb: Testing whether it significantly differs from 0 (determined by p.value).
r <.3 → weak association
.3 < r < .7 → moderate association
r >.7 → strong association
3. **Kendall’s t:** Similar to Spearman’s r, more common with small or high SD samples → rare.
108
Chi square x2: def
For nominal variables for testing whether it is significantly different from 0. Statistic measure of the difference between the observed and expected frequencies of the outcomes of a set of events or variables. Says something about the sample, not population!
Level of variables: Between 2 categorical (nominal/ordinal) variables.
109
Chi square x2: interpretation
Interpretation: Result HIGHER than 0 → correlation! Degrees of freedom and critical values, when the critical value (with p <.5 and df) is smaller than chi-square → reject H0.
Tables of critical values = All these values come together in tables of critical values for each test and you can see if your result is indeed generalizable. --> if the value of the chi-square is higher than the critical value, the more generalizable to the entire population. or: if the critical value is smaller than the chi-square, the null hypothesis must be rejected.
Critical value = DF against P in order to find the minimum threshold for the chi square to be representative for the whole population.
110
Simple linear regression
Describes the linear relationship between an explanatory variable and a response (x explains y). We predict the score on one variable (y) from the scores on another variable (x).
Level of variables: Between 2 continuous variables (interval/ratio).
111
Simple linear regression: statistical terms (4) + rule of thumb
Rule of thumb: b & β between -1 and 1 → same as correlation r.
b → Regression coefficient for the direction of the slope and shows the increase in x influences the value of y. Size of b says nothing about importance, because it is scale dependent.
β → Standardized coefficient (beta), compares the strength of the effect of each X on the Y. The larger the beta, the larger the effect.
R2 → Variance of y predicted by the model, how many y values can be predicted from x values → fitness of the model from 0 to 1. R2 of 0.9 is excellent, 0.7 pretty good, under 0.4 not that good.
hekje→ normal variables, no dummy.
112
Difference regression and pearson/spearman
The Pearson correlation gives information on the strength and direction of the linear relationship between two variables, while a simple linear regression analysis estimates parameters in a linear equation that can be used to predict values of one variable based on the other.
113
Multiple variables: types of analysis
1. Multiple linear regression: Multiple variables that predict Y.
Level of variables: Interval/ratio.
2. MANOVA:
Level of variables: More categorical independent variables and more continuous dependent variables.
114
Preconditions: multiple variables analysis
1. randomness of sample
2. normal distribution
115
Normal distribution + sample size rule of thumb
= Normal is with 50% of values on each side of the mean, and the two sides of the curve mirroring each other.
Sample size rule of thumb: 30+ minimum, 100-200 good size, 300-600 good size only if necessary
116
Normal distribution: z-score
Z-scores / standard scores: It describes the **individual score** above or below the mean, **and how far it is form the mean in SD**. Between -3 and 3. The closer to the 0 → close to the mean.
Z-score table: If z is 2.57, then the value is 0.49, meaning the values are in the 49 percent distance from the mean, which is at the extreme.
117
Studying macro effects: accounting: pros, cons, type
Accounting = Measurable, direct damage.
Pros → robust, comparison possible.
Cons → alternative explanations possible, no prospective estimates.
Time-series or within-country analysis = Directly counting the losses (e.g. educational, GDP etc.) due to conflict. Compares trends in outcome before, during, and after conflict.
118
Studying macro effects: counterfactual: pros, cons, types (2)
Scenario how economy/population etc. would have grown.
Pros → direct and indirect deaths included. Cons → relies on assumptions.
**Control or synthetic parameters**: Using extrapolation from prior parameters, from control region parameters, using a synthetic region.
**Comparison with region parameters**: E.g. neighboring countries. You can have *spillover effects*, however, if a comparable region is found, it allows to attribute observed changes in conflict.
119
Victim studies: limitations (3)
1. not having control groups, issue with **representativeness**.
2. Assessing crime rates and victimization → issue of **underreporting**.
3. Assessing impact of crime, post-crime life etc. → **ethical and causality issues**.
120
Indirect victimization: def and types (3)
= Indirect victims are those ‘who suffer harm as a result of the harm suffered by direct victims’ e.g. trauma, material losses etc.
Secondary victims = Psychological suffering caused by the death of a family member, material deprivation linked to the loss of the contributions brought by the deceased next of kin.
Other types: Communal indirect victimization effects (e.g. loss of cultural heritage etc.), intergenerational transmission of trauma.
121
Natural experiment
= Not all interventions are manipulated (e.g. child abusement). The impact of such events can only be studied non-experimentally in naturalistic, observational settings.
122
Analysis: Coding: process (3)
1. Preparing the data: Organising and structuring, writing memos etc.
2. First impressions: Little to no coding, only notes, initial thoughts on patterns.
3. Patterns & themes: Look at: (1) similarity, (2) difference, (3) frequency, (4) sequence, (5) correspondence (things happen in relation to other activities or events), (6) causation.
123
Deductive coding
= Pre-set coding scheme. Start with developing a codebook based on theory/prior research. Codebook is piloted and validated.
124
Inductive coding: def, types (3)
= No pre-set coding scheme. More iterative, **ground-up approach.** Derive codes from data. More exploratory.
Types:
Open coding = Break data into discrete parts → no theoretical coding!
Axial coding = Begin to draw connections, categorizing the open codes. Contextual connections, theoretical coding.
Hierarchical & selective coding = Connecting the categories, ‘axes’ in a network.
125
Emic / ethnography
Ethnography = Study that focuses on detailed and accurate description rather than explanations. Primary approach for anthropology, where you report on the subject’s term, emic approach (= from the respondent’s POV). Aims at understanding.
126
Etic / ethnomethodology
Ethnomethodology = Opposite of ethnography. It is rooted in ***phenomenology*** (= reality is socially constructed and researchers cannot rely on their subjects’ stories to depict social realities accurately).
127
Direct mortality
Military deaths + direct civilian deaths
128
Indirect mortality
= Indirect civilian deaths + civilian deaths post-conflict + military deaths post-conflict
129
Civilian mortality
= Direct civilian deaths + indirect civilian deaths + civilian deaths post-conflict
130
Military mortality
= Military deaths + military deaths post-conflict
131
Total human cost
All of the 5 types of deaths (direct, indirect civilian deaths, military deaths, post-conflict, civlian deaths post-conflict).
132
Emergency level mortality
2x CMR / U5MR → has it at least doubled?
133
Researching reactions: 2 focus
1. Functioning: What response mechanisms do → investigation, prosecution etc. and decision-making.
2. Consequences: Effects on (i) people and on (ii) rule of law/state-level. Like reconciliation, effects on victims and witnesses etc.
134
Existing datasets: issues (2)
1. Validity issues → existing data may not cover exactly what we are interested in, or the way we have conceptualized our variables.
2. Reliability issues → depends heavily on the quality of the datasets (completeness, over-underreporting, process of record-keeping).
135
Archival studies: limitations (3)
= Could be court archives e.g. appeals.
Limitations:
(a) missing data,
(b) often redactions,
© formal representation, no access to decision-making process (maybe corruption).
136
Direct observables
Physical charactersitics etc.
137
Indirect characteristics
Characteristics of a person as indicated by a person given in a self-admisnistered questionnaire.
138
Constructs
**Theoretical creations** that are based on observations but that cannot be observed (in)directly. A concept is, for instance, a construct.
139
Conceptualization
= The mental process whereby fuzzy and imprecise notions are made more specific and precise.
140
Concept
Constructs derived by **mutual agreement** from mental images (conceptions).
141
Cognitive interviewing
= Testing potential questions in an interview setting, probing to learn how respondents understand or interpret the questions.
142
Statistical significance
= A general term referring to the likelihood that relationships observed in a sample could be **attributed to sampling error alone**. Relationship is significant at the .05 level if the likelihood of its being only a function of sampling error is no greater than 5 out of 100.
143
Confidence levels
Measure of the percentage of test results that can be expected to be within a specified range. CI are frequently calculated at a confidence level of 95%. In general, if an observed result is statistically significant at a P-value of 0.05, then the null hypothesis should not fall within the 95% CI.
144
Type I error
= Refers to the incorrect rejection of the H0 → concluding that there is a relationship, when there actually is not a relationship in the population. Sampling error!
145
Type II error
= Refers to the incorrect acceptance of the H0 → concluding that there is no relationship, when there actually is a relationship in the population.
146
Design typology: mixed methods
A set of **different possible mixed methods designs** that attempts to convey the range of design options available for the use of mixed methods research.
147
Procedural diagram
A figure that depicts the **flow of the research activities** in a mixed methods study. Showing the quantitative and qualitative components and their stages in the study process, the research procedures within each stage, and the outcomes of each stage.
148
Strand: mixed methods
A component of a mixed methods study that encompasses the basic process of conducting quantitative or qualitative research.
149
What does a p value of .05 mean?
P-value indicates **whether the results are statistically significant**, and are **generalizable** to the population we draw the sample from. A p-value of .05 means that there is 95% percent sure (confidence interval) that there is a correlation. If p is bigger than .05 then are more errors.
